Wireless Networks L ecture 19: MIMO Peter Steenkiste CS and ECE, - PDF document

Wireless Networks L ecture 19: MIMO Peter Steenkiste CS and ECE, Carnegie Mellon University Peking University, Summer 2016 1 Peter A. Steenkiste Increasing Capacity: MIMO  Refresher: spatial diversity  Multiple-In Multiple-Out basics  MIMO in 802.11: » Single user MIMO: 802.11n » Multi user MIMO: 802.11ac 2 Peter A. Steenkiste Page 1

How Do We Increase Throughput in Wireless?  Wired world: Pull more wires!  Wireless world: How about if we could do the same thing and simply use more antennas? 3 Peter A. Steenkiste MIMO Multiple In Multiple Out N transmit M receive antennas antennas  N x M subchannels that can be used to send multiple data streams simultaneously  Fading on channels is largely independent » Assuming antennas are separate ½ wavelength or more  Combines ideas from spatial and time diversity, e.g. 1 x N and N x 1  Very effective if there is no direct line of sight » Subchannels become more independent 4 Peter A. Steenkiste Page 2

Why So Exciting? Method Capacity B log 2 (1 +  ) SISO B log 2 (1 +  ) Diversity (1xN or Nx1) B log 2 (1 +   ) Diversity (NxN) NB log 2 (1 +  ) Multiplexing 5 Peter A. Steenkiste Spatial Diversity  Use multiple antennas that pick up the signal in slightly different locations » Channels uncorrelated with sufficient antenna separation  Receiver diversity: i x H x P R = o h 1 y 1 y = h * x + n y x y =h * * (h * x + n) y 2 h 2  Receiver can pick strongest signal: y 1 or y 2  Or combines the signals: multiply y with the complex conjugate h * of the channel vector h » Can learn h based on training data 6 Peter A. Steenkiste Page 3

Other Diversity Options  Transmit diversity: i x P T x H = o x 1 h 1 x y x 2 h 2 i x P T x H x P R = o  Combined: h 11 y 1 x 1 h 12 x y h 21 x 2 y 2 h 22 7 Peter A. Steenkiste MIMO How Does it Work?  Transmit and receive multiple data streams  Coordinate the processing at the transmitter and receiver to overcome channel impairments » Maximize throughput or minimize interference T R I x P T x H x P R = O Precoding Channel Combining Matrix  Combines previous techniques 10 Peter A. Steenkiste Page 4

Mechanisms Supported by MIMO 11 Peter A. Steenkiste An Example of Space Coding 12 Peter A. Steenkiste Page 5

Direct-Mapped NxM MIMO M MxN N M Effect of transmission R = H * C + N Decoding O = P R * R C = I D DxM M N N Results O = P R * H * I + P R * N  How do we pick P R ? “Inverse” of H: H -1 » Equivalent of nulling the interfering possible (zero forcing) » Only possible if the paths are completely independent  Noise amplification is a concern if H is non- invertible – its determinant will be small » Minimum Mean Square Error detector balances two effects 14 Peter A. Steenkiste Precoded NxM MIMO M MxN N M Effect of transmission R = H * C + N Coding/decoding O = P R * R C = P T * I D DxM M N NxD D Results O = P R * H * P T * I + P R * N  How do we pick P R and P T ?  Singular value decomposition of H = U * S * V » U and V are unitary matrices – U H *U = V H *V = I » S is diagonal matrix 15 Peter A. Steenkiste Page 6

MIMO Discussion  Need channel matrix H: use training with known signal  So far we have ignored multi-path » Each channel is multiple paths with different properties » Becomes even messier!  MIMO is used in 802.11n » Can use two adjacent non-overlapping “WiFi channels” » Raises lots of compatibility issues » Potential throughputs of 100s of Mbps  Focus is on maximizing throughput between two nodes » Is this always the right goal? 17 Peter A. Steenkiste 802.11n Overview  802.11n extends 802.11 for MIMO » Supports up to 4x4 MIMO » Preamble that includes high throughput training field  Standardization was completed in Oct 2009, but, early products have long been available » WiFi alliance started certification based on the draft standard in mid-2007  Supported in both the 2.4 and 5 GHz bands » Goal: typical indoor rates of 100-200 Mbps; max 600 Mbps  Use either 1 or 2 non-overlapping channels » Uses either 20 or 40 MHz » 40 MHz can create interoperability problems  Supports frame aggregation to amortize overheads over multiple frames » Optimized version of 802.11e 18 Peter A. Steenkiste Page 7

802.11n Backwards Compatibility  802.11n can create interoperability problems for existing 802.11 devices (abg) » 802.11n does not sense their presence » Legacy devices end up deferring and dropping in rate  Mixes Mode Format protection embeds an n frame in a g or a frame » Preamble is structured so legacy systems can decode header, but MIMO can achieve higher speed (training, cod/mod info) » Works only for 20 MHz 802.11n use » Only deals with interoperability with a and g – still need CTS protection for b  For 40 MHz 802.11n, we need CTS protection on both the 20 MHz channels – similar to g vs. b » Can also use RTS/CTS (at legacy rates) » Amortize over multiple transmissions 19 Peter A. Steenkiste MIMO in a Network Context N transmit M receive antennas antennas How is this Different? M receive antennas N transmit - antennas M receivers 20 Peter A. Steenkiste Page 8

Multi-User MIMO Discussion  Math is similar to MIMO, except for the receiver processing (P R ) » Receivers do not have access to the signals received by antennas on other nodes » Limits their ability to cancel interference and extract a useful data stream » Closer to transmit MRC  MU-MIMO versus MIMO is really a tradeoff between TDMA and use of space diversity » Sequential short packets versus parallel long packets  Why not used in 802.11? 21 Peter A. Steenkiste Multi-User MIMO Up versus Down Link  Uplink: Multiple Access Channel (MAC) » Multiple clients transmit simultaneously to a single base station » Requires coordination among clients on packet transmission – hard problem because very fine-grained  Downlink: Broadcast Channel (BC) » Base station transmit separate data streams to multiple independent users » Easier to do: closer to traditional models of having each client receive a packet from the base station independently 22 Peter A. Steenkiste Page 9

802.11ac Multi-user MIMO  Extends beyond 802.11n » MIMO: up to 8 x 8 channels (vs. 4 x 4) » More bandwidth: up to 160 MHz by bonding up to 8 channels (vs. 40 MHz) » More aggressive signal coding: up to 256 QAM (vs. 64 QAM); both use 5/6 coding rate (data vs. total bits) » Uses RTS-CTS for clear channel assessment » Multi-gigabit rates (depends on configuration)  Support for multi-user MIMO on the downlink » Can support different frames to multiple clients at the same time » Especially useful for smaller devices, e.g., smartphones » Besides beam forming to target signal to device, requires also nulling to limit interference 23 Peter A. Steenkiste 802.11ad 60 GHz WiFi  Uses a new physical layer definition specifically for 60 GHz band » Very different signal propagation properties » Does not penetrate walls, but does work with reflections » Shorter distances » Small antennas and good beamforming properties  Defined up to 7 Gbps  Has been used for point-point links for a while » APs now available » Combined with other 802.11 versions » 802.11ad only available for short distances 24 Peter A. Steenkiste Page 10